[ocrfeeder] Fix the recognition for grayscale images

[Joaquim Manuel Pereira Rocha <jrocha src gnome org>]

commit 8793d5a550e2343405f3d3f88c3ff7ad1b2cf26e
Author: Joaquim Rocha <jrocha igalia com>
Date:   Fri Jul 22 20:53:49 2011 +0200

    Fix the recognition for grayscale images
    
    There were some errors when trying to recognize grayscale images
    and those had to do with the colors being integers instead of
    lists of integers.

 src/ocrfeeder/feeder/layoutAnalysis.py |    7 ++++++-
 src/ocrfeeder/util/graphics.py         |   31 ++++++++++++++-----------------
 2 files changed, 20 insertions(+), 18 deletions(-)
---
diff --git a/src/ocrfeeder/feeder/layoutAnalysis.py b/src/ocrfeeder/feeder/layoutAnalysis.py
index 344d923..13af038 100644
--- a/src/ocrfeeder/feeder/layoutAnalysis.py
+++ b/src/ocrfeeder/feeder/layoutAnalysis.py
@@ -511,7 +511,12 @@ class LayoutAnalysis(object):
             data_box.setText(text)
             data_box.setType(clip_type)
 
-        text_size = self.getTextSizeFromImage(clip, page_resolution)
+        if clip.mode == 'L':
+            grayscale_clip = clip
+        else:
+            grayscale_clip = clip.convert('L')
+        text_size = self.getTextSizeFromImage(grayscale_clip, page_resolution)
+
         if text_size:
             data_box.setFontSize(text_size)
 
diff --git a/src/ocrfeeder/util/graphics.py b/src/ocrfeeder/util/graphics.py
index 55bfdbf..fc2049a 100644
--- a/src/ocrfeeder/util/graphics.py
+++ b/src/ocrfeeder/util/graphics.py
@@ -93,22 +93,23 @@ def getTextSizeFromImage(image):
     background_color = 255
     if colors:
         colors.sort()
-        background_color = (list(colors[-1][1])).pop()
-    text_sizes = [0]
-    for i in xrange(getTextBeginHeight(image), height):
-        current_line = image.crop((0, i - 1, width, i))
-        current_line.convert('L')
+        background_color = colors[-1][1]
+    text_sizes = []
+    for i in xrange(1, height):
         blank_line = True
-        for i in range(0, current_line.size[0], 3):
-            color = current_line.getpixel((i, 0))
-            if colorsContrast(list(color).pop(), background_color):
+        for j in range(0, width, 3):
+            color = image.getpixel((j, i - 1))
+            if colorsContrast(color, background_color):
                 blank_line = False
                 break
         if blank_line:
-            if text_sizes[-1]:
+            if text_sizes and text_sizes[-1]:
                 text_sizes.append(0)
         else:
-            text_sizes[-1] += 1
+            if text_sizes and text_sizes[-1]:
+                text_sizes[-1] += 1
+            else:
+                text_sizes.append(1)
     text_sizes.sort()
     text_sizes = [i for i in text_sizes if i != 0]
     text_size = 0
@@ -127,13 +128,9 @@ def getTextBeginHeight(image):
     for i in range(1, height, 1):
         current_line = image.crop((0, i - 1, width, i))
         colors = current_line.getcolors()
-        foreground_color = [0, 0, 0]
-        if colors:
-            foreground_color = [0 for value in list((colors[0][1])[:-1])]
-        if colors:
-            for color in colors:
-                if list(color[1])[:-1] == foreground_color:
-                    return i
+        for color in colors:
+            if colorsContrast(color, 0) == 0:
+                return i
     return -1
 
 def getHorizontalAngleForText(image):